Shell Programming and Scripting

Hello again,

I am runnning a job on a cluster and I submit a job with the command

Code:

qsub script.sh

where script.sh is a script with the command i want o use. When i enter this command the cluster gives a huber like 123456 as the JOB ID.

I want the job to run about 12 hours and when it stops a new job will be submitted to the cluster. So what i do is creating a new script which includes the following:

Code:

for dir in $(seq 1 1 2)
do

msub script.sh

wait 123456 

done

but it fails waiting the job to finish and instead gives two jobs to the cluster.

Is it possible to do this?

Why not create a batch (*.bat) file such that once the first job finishes, the next one resumes?

thanks kayak for the answer but can you be more specific since I am new to this.. thank you

Please use code tags as required by forum rules!

If you hard code the number following wait, it will usually fail. Either use wait without a number, or find the process ID of your job and give this to wait.

man bash (not sure if you're using bash?):

wait is for shells with job control. Where jobs are local processes.
But "cluster" certainly means there a software that distributes the processes to other systems.
Then I can only recommend to consult the cluster software documentation that also describes the qsub and msub commands. There is certainly a "Queue-query" command...

Assuming this is referring to the qsub utility that is specified by the POSIX standards (although few implementations provide the optional POSIX Batch Environment Services and Utilities features), wait won't work. The qsub utility submits a job to be executed on a remote server and returns as soon as the job is queued on that server. Normally, the server running the job will send mail to the submitter when the job completes.

There are a few ways to do what I think you're trying to do. One way is to periodically run a qstat command with operands being the job IDs that the calls to the qsub utility returned for the jobs they submitted. You would then have to parse the qstat output to determine when a job completes and at that point kick off another qsub to replace the completed job on that server.

Another way would be to have the last step executed by the job you submit with qsub be another qsub command that will queue a job on the server that submitted the original qsub and have that job kick of a replacement if work remains to be done.